def test_multiple_test_correction(self):
pvalues = [0.001, 0.005, 0.1, 0.5, 0.01]
rejected_hypotheses = np.array([True, True, False, False, True])
corrected_pvalues = np.array([0.005, 0.0125, 0.125, 0.5, 0.01666667])
rej, cor = multiple_test_correction(pvalues)
self.assertTrue(np.array_equal(rej, rejected_hypotheses), msg="multiple_test_correction returns wrong list of "
"rejected hypotheses")
self.assertTrue(np.allclose(cor, corrected_pvalues), msg="multiple_test_correction returns wrong list of"
" corrected pvalues")
pvalues = [0.01, 0.05, 0.1, 0.5]
rejected_hypotheses = np.array([False, False, False, False])
corrected_pvalues = np.array([0.08333333, 0.20833333, 0.27777777, 1])
rej, cor = multiple_test_correction(pvalues, method="negcorr")
self.assertTrue(np.array_equal(rej, rejected_hypotheses), msg="multiple_test_correction(negcorr) returns wrong"
" list of rejected hypotheses")
self.assertTrue(np.allclose(cor, corrected_pvalues), msg="multiple_test_correction(negcorr) returns wrong"
" list of corrected pvalues")
with self.assertRaises(ValueError):
multiple_test_correction(pvalues, method="some_method")
def main():
# get command line arguments
args = parseArgs()
print(f"Using bam: {args.bam}")
bf = args.bam
print(f"Will write peaks: {args.outfile}_peaks.tsv")
of = args.outfile
cf = args.controlfile
cs = args.chromsizes
pvalue = args.pvalue
minreads = args.minreads
minsize = args.minsize
bs = args.binsize
corr = args.correct_pval
if corr not in ["bh", "by", None]:
print("Invalid correction method (please pass either 'bh' or 'by'")
sys.exit(1)
res, cov, cov2 = call_peaks(bf, cs, pvalue, minreads, bs, cfile=cf)
# rpm norm the signal before writing to bigwig
cov2.coverage = np.array(cov2.coverage,
dtype='object') * (1e6 / float(cov2.reads))
bwfile = of + ".bw"
write_bigwig(cov2, bwfile, cs)
# write peaks to file
outbed = of + "_peaks.tsv"
write_bed(res, outbed, minsize)
if corr is not None:
dat = pd.read_csv(
outbed,
sep="\t",
names=["chr", "start", "end", "name", "score", "strand"])
if corr == "bh":
b, corr = multiple_test_correction(dat.score, method="p")
elif corr == "by":
b, corr = b, corr = multiple_test_correction(dat.score, method="n")
dat["score"] = corr
dat["score"] = dat["score"].apply(lambda x: -np.log(x))
dat[dat.score > -np.log10(pvalue)].to_csv(outbed,
sep="\t",
header=False,
index=False)
write_counts(of, dat)
def filter_by_pvalue_strand_lag(ratios, pcutoff, pvalues, output, no_correction, name, singlestrand):
"""Filter DPs by strang lag and pvalue"""
if not singlestrand:
zscore_ratios = zscore(ratios)
ratios_pass = np.where(np.bitwise_and(zscore_ratios > -2, zscore_ratios < 2) == True, True, False)
if not no_correction:
pv_pass = [True] * len(pvalues)
pvalues = map(lambda x: 10**-x, pvalues)
_output_BED(name + '-uncor', output, pvalues, pv_pass)
_output_narrowPeak(name + '-uncor', output, pvalues, pv_pass)
pv_pass, pvalues = multiple_test_correction(pvalues, alpha=pcutoff)
else:
pv_pass = np.where(np.asarray(pvalues) >= -log10(pcutoff), True, False)
if not singlestrand:
filter_pass = np.bitwise_and(ratios_pass, pv_pass)
assert len(pv_pass) == len(ratios_pass)
else:
filter_pass = pv_pass
assert len(output) == len(pvalues)
assert len(filter_pass) == len(pvalues)
return output, pvalues, filter_pass
def filter_by_pvalue_strand_lag(ratios, pcutoff, pvalues, output,
no_correction, name, singlestrand):
"""Filter DPs by strang lag and pvalue"""
if not singlestrand:
zscore_ratios = zscore(ratios)
ratios_pass = np.where(
np.bitwise_and(zscore_ratios > -2, zscore_ratios < 2) == True,
True, False)
if not no_correction:
pv_pass = [True] * len(pvalues)
pvalues = map(lambda x: 10**-x, pvalues)
_output_BED(name + '-uncor', output, pvalues, pv_pass)
_output_narrowPeak(name + '-uncor', output, pvalues, pv_pass)
pv_pass, pvalues = multiple_test_correction(pvalues, alpha=pcutoff)
else:
pv_pass = np.where(np.asarray(pvalues) >= -log10(pcutoff), True, False)
if not singlestrand:
filter_pass = np.bitwise_and(ratios_pass, pv_pass)
assert len(pv_pass) == len(ratios_pass)
else:
filter_pass = pv_pass
assert len(output) == len(pvalues)
assert len(filter_pass) == len(pvalues)
return output, pvalues, filter_pass
def multiple_correction(dic):
"""
dic[ty][r][q] = p
"""
for ty in dic.keys():
all_p = []
rn = len(dic[ty].keys())
qn = len(dic[ty].values()[0].keys())
cue = {}
i = 0
if rn == 1 and qn == 1: return
# get all p values from the dictionary
for r in dic[ty].keys():
for q in dic[ty][r].keys():
if isinstance(dic[ty][r][q], str):
pass
else:
all_p.append(dic[ty][r][q])
cue[ty + r + q] = i
i = i + 1
# correction
reject, pvals_corrected = multiple_test_correction(all_p, alpha=0.05, method='indep')
# modify all p values
for ir, r in enumerate(dic[ty].keys()):
for iq, q in enumerate(dic[ty][r].keys()):
try:
dic[ty][r][q] = pvals_corrected[cue[ty + r + q]]
except:
pass
def multiple_correction(dic):
"""
dic[ty][r][q] = p
"""
for ty in dic.keys():
all_p = []
rn = len(dic[ty].keys())
qn = len(dic[ty].values()[0].keys())
cue = {}
i = 0
if rn == 1 and qn == 1: return
# get all p values from the dictionary
for r in dic[ty].keys():
for q in dic[ty][r].keys():
if isinstance(dic[ty][r][q], str):
pass
else:
all_p.append(dic[ty][r][q])
cue[ty + r + q] = i
i = i + 1
# correction
reject, pvals_corrected = multiple_test_correction(all_p,
alpha=0.05,
method='indep')
# modify all p values
for ir, r in enumerate(dic[ty].keys()):
for iq, q in enumerate(dic[ty][r].keys()):
try:
dic[ty][r][q] = pvals_corrected[cue[ty + r + q]]
except:
pass
def test_multiple_test_correction_using_R(self):
pvalues_list = [[0.001, 0.005, 0.1, 0.5, 0.01], [0.03, 0.8, 0.47,0.1], [0.0003, 0.4, 0.002]]
# corrected pvalues calculated by using R function p.adjust
corrected_pvalues = [np.array([0.005, 0.0125, 0.125, 0.5, 0.01666667]), np.array([0.12, 0.8, 0.62666667, 0.2]),
np.array([0.0009, 0.4, 0.003])]
# get corrected pvalues from own function
for i in range(0, len(pvalues_list)):
res = multiple_test_correction(pvalues_list[i])
self.assertTrue(np.allclose(res[1], corrected_pvalues[i]), msg="multiple_test_correction returns wrong list"
"of corrected pvalues")
def generate_rna_exp_pv_table(root, multi_corr=True):
"Generate p value table for Experiments vs RNA in the same project"
nested_dict = lambda: defaultdict(nested_dict)
#nested_dict = lambda: defaultdict(lambda: 'n.a.')
data = nested_dict()
rnas = []
for item in os.listdir(root):
pro = os.path.join(root, item, "profile.txt")
if os.path.isfile(pro):
with open(pro) as f:
for line in f:
if line.startswith("Experiment"): continue
else:
line = line.strip().split("\t")
data[item][line[0]] = float(line[7])
rnas.append(line[0])
exp_list = sorted(data.keys())
rnas = sorted(list(set(rnas)))
pvs = []
for rna in rnas:
for exp in exp_list:
if data[exp][rna]: pvs.append(data[exp][rna])
reject, pvals_corrected = multiple_test_correction(pvs, alpha=0.05, method='indep')
with open(os.path.join(root, "table_exp_rna_pv.txt"), "w") as t:
print("\t".join(["RNA_ID"] + exp_list), file=t)
i = 0
for rna in rnas:
newline = [rna]
for exp in exp_list:
if data[exp][rna]:
newline.append(str(pvals_corrected[i]))
i += 1
else:
newline.append("n.a.")
print("\t".join(newline), file=t)
for d, p in plist.iteritems():
list_all_index(path=os.path.dirname(p),
link_d=dirlist, show_RNA_ass_gene=show_RNA_ass_gene)
def generate_rna_exp_pv_table(root, multi_corr=True):
"Generate p value table for Experiments vs RNA in the same project"
nested_dict = lambda: defaultdict(nested_dict)
# nested_dict = lambda: defaultdict(lambda: 'n.a.')
data = nested_dict()
rnas = []
for item in os.listdir(root):
pro = os.path.join(root, item, "profile.txt")
if os.path.isfile(pro):
with open(pro) as f:
for line in f:
if line.startswith("Experiment"):
continue
else:
line = line.strip().split("\t")
data[item][line[0]] = float(line[7])
rnas.append(line[0])
exp_list = sorted(data.keys())
rnas = sorted(list(set(rnas)))
pvs = []
for rna in rnas:
for exp in exp_list:
if data[exp][rna]: pvs.append(data[exp][rna])
if multi_corr:
reject, pvals_corrected = multiple_test_correction(pvs, alpha=0.05, method='indep')
else:
pvals_corrected = pvs
with open(os.path.join(root, "table_exp_rna_pv.txt"), "w") as t:
print("\t".join(["RNA_ID"] + exp_list), file=t)
i = 0
for rna in rnas:
newline = [rna]
for exp in exp_list:
if data[exp][rna]:
newline.append(str(pvals_corrected[i]))
i += 1
else:
newline.append("n.a.")
print("\t".join(newline), file=t)
def get_peaks(name, DCS, states, ext_size, merge, distr, pcutoff,
no_correction):
indices_of_interest = DCS.indices_of_interest
first_overall_coverage = DCS.first_overall_coverage
second_overall_coverage = DCS.second_overall_coverage
c1 = list(first_overall_coverage)
c2 = list(second_overall_coverage)
tmp_peaks = []
for i in range(len(indices_of_interest)):
if states[i] not in [1, 2]:
continue #ignore background states
strand = '+' if states[i] == 1 else '-'
cov1 = c1[indices_of_interest[i]]
cov2 = c2[indices_of_interest[i]]
chrom, start, end = DCS._index2coordinates(indices_of_interest[i])
tmp_peaks.append((chrom, start, end, cov1, cov2, strand))
i, j = 0, 0
peaks = []
pvalues = []
while i < len(tmp_peaks):
j += 1
c, s, e, c1, c2, strand = tmp_peaks[i]
v1 = [c1]
v2 = [c2]
#merge bins
while i + 1 < len(tmp_peaks) and e == tmp_peaks[
i + 1][1] and strand == tmp_peaks[i + 1][5]:
e = tmp_peaks[i + 1][2]
v1.append(tmp_peaks[i + 1][3])
v2.append(tmp_peaks[i + 1][4])
i += 1
s1 = sum(v1)
s2 = sum(v2)
if s1 + s2 > SIGNAL_CUTOFF:
pvalues.append(('NA', 'NA', 'NA', 'NA'))
else:
if strand == '+':
pvalues.append((s1, s2, 'l', distr))
else:
pvalues.append((s1, s2, 'r', distr))
peaks.append((c, s, e, s1, s2, strand))
i += 1
print('Number of Peaks where p-value is not calculated: ',
pvalues.count(('NA', 'NA', 'NA', 'NA')),
file=sys.stderr)
#pool = multiprocessing.Pool(processes=2)#multiprocessing.cpu_count() * 3/2)
pvalues = map(_compute_pvalue, pvalues)
assert len(pvalues) == len(peaks)
merge_delete(ext_size, merge, peaks, pvalues, name)
#peaks = [(c, s, e, s1, s2, strand)]
if not no_correction:
pvalues = map(lambda x: 10**-x, pvalues)
pv_pass, pvalues = multiple_test_correction(pvalues, alpha=pcutoff)
pvalues = map(_get_log10pvalue, pvalues)
else:
pv_pass = [True] * len(pvalues)
_output_BED(name, pvalues, peaks, pv_pass)
_output_narrowPeak(name, pvalues, peaks, pv_pass)
def get_peaks(name, DCS, states, ext_size, merge, distr, pcutoff, no_correction):
indices_of_interest = DCS.indices_of_interest
first_overall_coverage = DCS.first_overall_coverage
second_overall_coverage = DCS.second_overall_coverage
c1 = list(first_overall_coverage)
c2 = list(second_overall_coverage)
tmp_peaks = []
for i in range(len(indices_of_interest)):
if states[i] not in [1,2]:
continue #ignore background states
strand = '+' if states[i] == 1 else '-'
cov1 = c1[indices_of_interest[i]]
cov2 = c2[indices_of_interest[i]]
chrom, start, end = DCS._index2coordinates(indices_of_interest[i])
tmp_peaks.append((chrom, start, end, cov1, cov2, strand))
i, j = 0, 0
peaks = []
pvalues = []
while i < len(tmp_peaks):
j+=1
c, s, e, c1, c2, strand = tmp_peaks[i]
v1 = [c1]
v2 = [c2]
#merge bins
while i+1 < len(tmp_peaks) and e == tmp_peaks[i+1][1] and strand == tmp_peaks[i+1][5]:
e = tmp_peaks[i+1][2]
v1.append(tmp_peaks[i+1][3])
v2.append(tmp_peaks[i+1][4])
i += 1
s1 = sum(v1)
s2 = sum(v2)
if s1 + s2 > SIGNAL_CUTOFF:
pvalues.append(('NA', 'NA', 'NA', 'NA'))
else:
if strand == '+':
pvalues.append((s1, s2, 'l', distr))
else:
pvalues.append((s1, s2, 'r', distr))
peaks.append((c, s, e, s1, s2, strand))
i += 1
print('Number of Peaks where p-value is not calculated: ', pvalues.count(('NA', 'NA', 'NA', 'NA')), file=sys.stderr)
#pool = multiprocessing.Pool(processes=2)#multiprocessing.cpu_count() * 3/2)
pvalues = map(_compute_pvalue, pvalues)
assert len(pvalues) == len(peaks)
merge_delete(ext_size, merge, peaks, pvalues, name)
#peaks = [(c, s, e, s1, s2, strand)]
if not no_correction:
#first output uncorrected p-values
pv_pass = [True] * len(pvalues)
_output_BED(name + '-uncor', pvalues, peaks, pv_pass)
_output_narrowPeak(name + '-uncor', pvalues, peaks, pv_pass)
#then correct p-values and output
pvalues = map(lambda x: 10**-x, pvalues)
pv_pass, pvalues = multiple_test_correction(pvalues, alpha=pcutoff)
pvalues = map(_get_log10pvalue, pvalues)
else:
pv_pass = list(np.where(np.asarray(pvalues) >= -log10(pcutoff), True, False))
_output_BED(name, pvalues, peaks, pv_pass)
_output_narrowPeak(name, pvalues, peaks, pv_pass)